Section checkpoint

Data First

9 questions drawn from the 3 chapters in this section. If most land on the first try, you are ready for the next section. If not, the explanation after each answer says which idea to revisit.

Checkpoint

9 questions · first-try answers earn the most XP

1. Why might declining usage be a better churn feature than raw usage?

2. What is the difference between features and labels?

3. A feature turns out to be a proxy for the label in disguise. What should you do?

4. You predict next month's churn from six months of history. How should you split?

5. The same customer appears twice, once in train and once in test. What is the consequence?

6. Why fix `random_state` when splitting and fitting?

7. When can deleting rows with missing values be acceptable?

8. A categorical column has 300 distinct city names. Why does one-hot encoding become a problem?

9. Why must imputers, encoders and scalers be fitted on the training split only?

What this checkpoint covers

Next section: Core Models